Title: Mapping snow cover using MODIS Part I: The MODIS Instrument Part II: Normalized Difference Snow Index Part III: Quality Control Procedures and Masks Part IV: Apply masks to create a corrected snow map Product Type: Curriculum Developer: Helen Cox (Professor, Geography, California State University, Northridge): [email protected]Maziyar Boustani & Laura Yetter (Research Assts., Institute for Sustainability, California State University, Northridge) Target audience: Undergraduate Format: Tutorial (pdf document) Software requirements * : ArcMap 9 or higher (ArcGIS Desktop) (Parts II, IV), ERDAS Imagine 2010 or higher (Parts I, II, III, IV) Data: All data required are obtained within the exercise. Estimated time to complete: All parts: 7 hrs. Part I: 2 hrs. Part II: 2 hrs. Part III: 2 hrs. Part IV: 1 hr. Alternative Implementations: • Parts I and II together provide a standalone exercise producing a snow map • Parts I, II and Part IV (starting at #2) together provide a standalone exercise producing a snow map and comparing it to one produced by NASA • Completing all parts (I through IV) produce a snow map with corrections that is compared to one produced by NASA Learning objectives: Part I: • Learn about the MODIS instrument and MODIS data • Download MODIS data Part II: • Learn about the Normalized Difference Snow Index (NDSI) • Create a Model in ERDAS Imagine to calculate the NDSI • Create a snow map Part III: • Learn how to identify water and forests where snow could be misidentified • Create a Normalized Difference Vegetation Index (NDVI) image • Create masks that will be used to eliminate water and dark forests from the NDSI Part IV: • Apply the water, forest, and NDVI masks to eliminate water and forest from the snow map • Re-project the snow map and compare to a MODIS snow product map * Tutorials may work with earlier versions of software but have not been tested on them
13
Embed
Title: Mapping snow cover using MODIS Mapping snow cover using MODIS ... ERDAS Imagine 2010 or higher ... • Create a Normalized Difference Vegetation Index ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Title: Mapping snow cover using MODIS Part I: The MODIS Instrument Part II: Normalized Difference Snow Index Part III: Quality Control Procedures and Masks Part IV: Apply masks to create a corrected snow map Product Type: Curriculum Developer: Helen Cox (Professor, Geography, California State University,
Northridge): [email protected] Maziyar Boustani & Laura Yetter (Research Assts., Institute for Sustainability, California State University, Northridge)
Target audience: Undergraduate Format: Tutorial (pdf document) Software requirements*: ArcMap 9 or higher (ArcGIS Desktop) (Parts II, IV),
ERDAS Imagine 2010 or higher (Parts I, II, III, IV) Data: All data required are obtained within the exercise. Estimated time to complete: All parts: 7 hrs. Part I: 2 hrs. Part II: 2 hrs. Part III: 2 hrs. Part IV: 1 hr. Alternative Implementations: • Parts I and II together provide a standalone exercise producing a
snow map • Parts I, II and Part IV (starting at #2) together provide a standalone
exercise producing a snow map and comparing it to one produced by NASA
• Completing all parts (I through IV) produce a snow map with corrections that is compared to one produced by NASA
Learning objectives: Part I: • Learn about the MODIS instrument and MODIS data • Download MODIS data
Part II: • Learn about the Normalized Difference Snow Index (NDSI) • Create a Model in ERDAS Imagine to calculate the NDSI • Create a snow map
Part III: • Learn how to identify water and forests where snow could be
misidentified • Create a Normalized Difference Vegetation Index (NDVI) image • Create masks that will be used to eliminate water and dark forests
from the NDSI Part IV:
• Apply the water, forest, and NDVI masks to eliminate water and forest from the snow map
• Re-project the snow map and compare to a MODIS snow product map
*Tutorials may work with earlier versions of software but have not been tested on them
Mapping snow cover using MODIS
Part II: Normalized Difference Snow Index
Objectives
Learn about the Normalized Difference Snow Index (NDSI)
Create a Model in ERDAS Imagine to calculate the NDSI
Create a snow map
NDSI
Information on snow cover is important for a variety of reasons (estimates of freshwater storage,
indicator of global change, climate modeling and feedback effects). Trying to identify snow cover only
using the visible reflected light may be difficult because many things appear white such as clouds, sand,
surf, salt beds or even rocks. MODIS is used to monitor snow cover from space using the Normalize
Difference Snow Index (NDSI), with some corrections added for special circumstances. The NDSI uses
visible and short wave near infrared bands to identify snow.
Researchers have studied the spectral reflectance of snow (see figure below). It has a high reflectance in
band 4 (0.545-0.565 µm, visible green) of the MODIS instrument and a low reflectance in band 6 (1.628-
1.652 µm, short wavelength near infrared).
500m Band Number Wavelength (nm) Color
Sur_refl_b01 0.620- 0.670 Red
Sur_refl_b02 0.841- 0.876 NIR
Sur_refl_b03 0.459- 0.479 Blue
Sur_refl_b04 0.545- 0.565 Green
Sur_refl_b05 1.230- 1.250 NIR
Sur_refl_b06 1.628- 1.652 MIR
Sur_refl_b07 2.105- 2.155 MIR
The NDSI is a normalized ratio of the difference in reflectance in these bands that takes advantage of the
unique signature and spectral differences to identify snow from surrounding features even clouds. The
equation for the NDSI is
NDSI= (band 4 – band 6)/( band 4 + band 6)
NDSI is calculated on a pixel-by-pixel basis and will generate a gray scale image with high values (bright
pixels) representing snow. The NDSI not only distinguishes snow versus non snow covered areas and
clouds, but also reduces the influence of atmospheric effects in the readings. The MODIS algorithm sets
a threshold value of 0.4 for snow (i.e. if NDSI > 0.4 pixel is snow, else not snow).
There are some nuances, however, with distinguishing snow from other non snow features such as
water or dense forests because they have similar NDSI readings to snow. These features have low
reflectance (they are good absorbers) and cause the NDSI denominator to be small. Then only small
increases in band 4 are enough to increase the NDSI value and misclassify a pixel as snow (Hall et al
2002). Therefore in addition to the NDSI it is necessary to examine reflectance at other wavelengths to
distinguish these features from snow cover.
To separate water bodies from snow, a pixel should be mapped as water if the reflectance in band 2
(0.841- 0.876 µm) is less than 11% even if the NDSI is greater than or equal to 0.4. To identify dark
forested areas, if the NDSI is greater than or equal to 0.4 (snow) but the reflectance in band 4 (0.545 -
0.565 µm) is less than 10% a pixel will be marked as forest. If a pixel NDSI value is less than 0.4 (not
snow) but the NDVI is approximately 0.1 the pixel could be snow-covered forest (Hall et al 2002). These
measures prevent low reflective features like water bodies and dense forest canopies from being
misclassified as snow. These corrections will be addressed in future exercises.
NDSI Model
In this exercise we will build a model in ERDAS Imagine to calculate NDSI from the MODIS image we
downloaded in the last exercise. Note that the pixel values in the reflectance image are whole numbers
with a range from -100 to 16000 and pixels of no data have a value of -28672. In order to calculate NDSI
they need to be scaled by a factor of 0.0001. See documentation at: